Parole decisions about perpetrators of domestic violence in England and Wales

This study, the first to examine parole decisions for per-petrators of domestic violence in England and Wales, examines how parole boards reach decisions through a thematic analysis of interviews with 20 Parole Board members, alongside logistic regressions of all 137 parole decisions about perpetrators in an 18-month period. In this sample, recommendations of professionals, especially psychologists, were by far the strongest predictor of a release decision. The nature of offending was also significant in predicting release. Size and composition of the panel may also affect decisions. More panel members, and longitudinal feedback, may encourage more reflective, considered decision making.


INTRODUCTION
The intense physical and psychological harm caused by perpetrators of domestic violence affects over one in five adults (Office for National Statistics, 2022) and children (Radford et al., 2011) in England and Wales over their lifetime.Domestic violence is both a crime in itself and a contributing factor to crime among its victims -female offenders overwhelmingly suffer domestic violence which derails their own rehabilitation (Prison Reform Trust, 2017).When the most serious perpetrators of domestic violence are jailed, over 20% in the UK go on to reoffend within two years of release (Bloomfield & Dixon, 2015), similar to patterns observed internationally (McKay, 2022).Domestic violence is therefore a focus in current criminal justice policy in England and Wales, including legislation that recognises children exposed to domestic violence as direct victims of that abuse (Domestic Abuse Act 2021) and enhanced electronic monitoring of offenders after release to protect victims (Siddique, 2023).
Given the complex risks posed by perpetrators after release, it is important to consider how the decisions to release them from prison are made, as well as risk management after release.When parole boards evaluate risk that principally involves the danger of harming an intimate partner, they analyse a highly gendered harm perpetrated within homes and intimate settings.The recent politicisation of parole decisions, and the focus in criminal justice on violence against women and girls, add further layers to this decision making.These considerations further demonstrate the importance of studying parole decisions for domestic violence perpetrators.
This research is the first to focus on parole decisions for domestic violence perpetrators, exploring mechanisms behind those decisions as well as factors associated with a decision to release an offender.It is timely, since the parole decision-making process has recently been subject to contentious changes to the permitted use of expert recommendations, subsequently ruled unlawful (R (Bailey and Morris) v. Secretary of State for Justice (2023)).
This article reports research on Parole Board decision making regarding domestic violence perpetrators.Before describing the research, we explore in some detail why parole decision making for that subset of perpetrators might differ subtly from parole decisions for crimes in general.
In so doing we summarise the existing literature on parole decisions and decision making more widely, outlining the cultural and political context of parole in England and Wales.This provides the context by which to understand the report of the research itself, which follows this overview.

The politicisation of parole decisions
When it started work in 1968, the Parole Board of England and Wales advised the Home Secretary on whether prisoners could be released after serving one-third of their term.Initially, their goals were framed in terms of reducing the prison population and acknowledging the rehabilitation of prisoners (Padfield, 2019).This role has evolved to place a greater emphasis on managing risk, and in the early 2020s, their decisions have become increasingly politicised.Both major political parties used parole decisions as an opportunity to position themselves on the side of 'the people' and demonstrate the 'toughness' of the state (Annison & Guiney, 2022).This tendency increased after the public uproar over the decision to release serial rapist, John Worboys, in 2018 (Annison, 2020), a decision overturned when the High Court ruled that the Parole Board should have made 'further inquiry' into the offences for which he was not convicted.This creates a fraught environment for Parole Board members aiming to use professional expertise to make a balanced decision about risks.
The past decade has also seen an attempt to privatise the system for monitoring prisoners after release, subsequently reversed but still affecting the perceived effectiveness of the probation service (Albertson & Corcoran, 2020).

2.2
The shift to risk management The Parole Board in England and Wales receives applications from prisoners seeking to be released on licence or moved to an 'open' prison (where they spend time in the community preparing for release).
The test for release is laid out in formal guidance: The Parole Board is empowered to direct release if it is satisfied that it is no longer necessary for the protection of the public that they be confined.(Parole Board, 2019, p.8)However, the Parole Board's role suffers a 'central crisis' (Hood & Shute, 2000): it is impossible to predict someone's future behaviour, and pre-emptively detaining someone for a potential future amounts to re-trying or 'resentencing' an offender (Dagan, 2021) for their original offences.This reflects two shifts in the modern history of parole in England and Wales: a shift from rehabilitation to risk management and retribution (Guiney, 2022); and a politicisation of Parole Board decisions (Ministry of Justice, 2022).These shifts occur despite a growing prison population with growing financial and public health costs (Baybutt, Dooris & Farrier, 2019).Meanwhile, experts recommend greater independence in the parole process, greater emphasis on findings-of-fact, and a renewed focus on rehabilitation (JUSTICE, 2022).

The extension of parole decision making into indeterminate sentences
Since 2003, the Parole Board of England and Wales has seen its role shift from making decisions about prisoners subject to fixed-term sentences (Padfield, 2019) to a greater focus on indeterminate sentences without a fixed end date.
Imprisonment for Public Protection (IPP) sentences set a minimum sentence, after which the prisoner must prove that they are suitable for release.These were abolished in 2012 (leaving life sentences as the only form of indeterminate sentence that a judge can hand down) but, as of 2022, 2,926 prisoners remained in prison indefinitely, long past their minimum tariff (House of Commons Justice Committee, 2022).Almost half of these had been recalled to custody after release.These offenders often present a continued risk of offending, and are more likely than the average prisoner to experience mental illness and substance addiction (McConnell & Raikes, 2019;Rutherford, 2009), but there is pressure on the Parole Board to move such prisoners out of a penal system which may exacerbate those problems (Annison & Condry, 2022;Straub & Annison, 2020).
Prisoners subject to either a life sentence or IPP sentence are reliant on meeting the Parole Board if they are ever to be released (Padfield, Liebling & Arnold, 2000) -prisoners on life sentences can be released by the Parole Board but remain on licence for the rest of their lives.This focus on indeterminate sentences departs from the role of the Parole Board before the 1990s, which was mostly to rule on early release for prisoners on determinate sentences (Padfield, 2019).

How parole boards decide to release prisoners
The parole decision involves interactions between different levels (Proctor, 1999): organisational factors such as professional recommendations, and the type and length of sentence; and casespecific characteristics.Professional recommendations serve as a 'sponsor' for a release decision, prior to discussing what Meyer (2001) refers to as 'magic' characteristics -aspects of a prisoner's custodial behaviour, or how they come across in person, that satisfy the board (often at an intuitive level) that the prisoner can be released.In practice, this means that Parole Board members often incorporate structural variables and previous recommendations, before weighing up the case-specific risks involved.

Offender specific considerations for parole boards
Parole decision making incorporates both the offender's inherent ability to reduce their own risk of reoffending, and how the offender can be managed in the community (Bottoms, 2001).
Existing research highlights that a prisoner is more likely to be released if: (i) They have impressed their offender supervisors with their 'institutional behaviour' (Caplan, 2010;Huebner & Bynum, 2008); (ii) they are 'lucky' enough (Padfield, 2019) to have an offender manager (probation officer) who has satisfied the board that their risk management plan is robust (Bradford & Cowell, 2012); (iii) they have a stable family and community ties (Mooney & Daffern, 2014); (iv) they have attended various kinds of perpetrator programmes (for all categories of offender) designed to reduce their risk of offending (Caplan, 2010); (v) they have reduced their alcohol and drug use (Bradford & Cowell, 2012).

Wider factors in decision making
The individual, offender-specific factors sit within wider institutional factors that frame the Parole Board's decision making.There is tension between gauging 'societal risk' (the risk that an offender will inflict harm after release) and 'organisational risk' (the risk that the organisation will be damaged as a result of their decision) -the two overlap but differ, and while Parole Board members are required to address societal risk, like any members of an organisation with statutory responsibilities (especially one making such sensitive, high-stakes decisions) they also feel the pressure of organisational risk (Kemshall, 2002).Conscious decision making for an individual case only makes up part of the decision, which takes place within overlapping structures including wider societies and the Parole Board members' own professional and personal backgrounds (Huebner & Bynum, 2008).Unconscious influences have been identified in empirical analysis of criminal justice decisions (Dhami & Belton, 2017) but not yet in parole research.
The nature of the parole decision is susceptible to 'representativeness heuristics' (Kahneman, Knetsch & Thaler, 1991), where we assign someone to a category based on our sense of whether they resemble other people in that category.The Parole Board member, trying to predict future offending, aims to assign a prisoner to the category of 'people likely to reoffend' or 'people unlikely to reoffend' based on limited information, drawing on their experience of people in each category.Parole Board members may also experience the 'status quo bias' (Kahneman, Knetsch & Thaler, 1991), where the default option is to avoid changing the situation -in this case, a status quo bias would favour keeping the prisoner incarcerated.
These shortcuts are more significant when the decision maker is acting, as parole boards tend to (Padfield, Liebling & Arnold, 2000), on 'intuition'.However, this does not necessarily reduce the accuracy of their understanding of risk (Kahneman & Klein, 2009), especially when decisions are 'high-validity', where the decision maker gets prompt feedback on decisions which are frequent and regular in nature.

Decisions about perpetrators of domestic violence
Domestic violence is distinct from crimes in general in custodial considerations: it is overwhelmingly a male-on-female offence, and the role of control and intimidation frames the way in which we need to assess risk (Williams & Walklate, 2020).The increase in domestic violence during the coronavirus pandemic from 2020 (Piquero et al., 2021) increases the pressure to address this kind of offending.Domestic violence includes physical violence and non-physical forms of intimidation, coercion and control (Dobash & Dobash, 2004;Wiener, 2017) and involves violence towards a current or former intimate partner regardless of gender or marital status.Its intimate nature means that up until the end of the last century it was treated as private and outside the jurisdiction of the criminal justice system.Therefore, the use of the criminal justice system to tackle this type of offence is relatively recent historically (Williams & Walklate, 2020) and complicated by the nature of most domestic violence as an ongoing aspect of a couple's life together and the degree to which the perpetrator exerts control over their victim (Johnson, 2006) .This distinguishes it from the easily-compartmentalised 'incidents' that are normally seen in criminal justice (Kelly & Westmarland, 2016).

4.2
The gendered nature of domestic violence Domestic violence is one of the most gendered crimes, alongside rape and sexual offences (which often feature within domestic violence).The gendered nature of domestic violence has been subject to debate.Some surveys report equivalence in men and women's perpetration of violence (Straus & Gozjolko, 2016) while others (based on similar surveys and professional reports) find far greater harm perpetrated by men against women than vice versa (Razera, Gaspodini & Falcke, 2017).The discrepancy is explained by the differences in methodologies (Dobash & Dobash, 2004): when counting number of incidents with male and female perpetrators, the frequencies are comparable if not equal (Straus & Gozjolko, 2016), but when distinguished by 'nature of offending', a distinction emerges (Johnson, 2006).'Situational couple violence' where the perpetrator harms but does not control their victim, is perpetrated by women at a comparable frequency as by men, but 'intimate terrorism', where the perpetrator exerts control and intimidates the victim, is overwhelmingly perpetrated by men (Johnson, 2006).This reflects the greater overall harm, severity of injuries, and frequency of murders (Kimmel, 2002) committed by men towards female partners than vice versa.The gendered nature of offending matters when considering the importance of reconceptualising prisoners' sense of masculinity and self-image as a man interacting with society, and especially with women (Maguire, 2021;Ricciardelli, Maier & Hannah-Moffat, 2015).It therefore matters to the Parole Board's evaluation of the specific risks posed by this group of offenders, as the board explores the way in which the offender's identity and attitudes towards women have changed since their offending.The legal test for all offenders is the same, but parole boards benefit from drawing on offence-specific research.
Parole boards faced with domestic violence perpetrators need to use a different approach to evaluating risk posed by more 'chaotic' offenders, than the approach used for offenders who control their victims, based on the extensive research into predicting domestic violence reoffending.Perpetrators are more likely to reoffend when they have committed offences of a high frequency and severity (McKay, 2022), and when their offences have included controlling elements akin to Johnson's 'intimate terrorism' (Wiener, 2017).
Once convicted, offenders often attend domestic violence perpetrator programmes (DVPPs), designed to reduce their risk of reoffending, but the effectiveness of these programmes is questionable (Haggård et al., 2017;Vigurs et al., 2016).Likewise, offenders who reduce their drug and alcohol use are not predicted to reduce their offending accordingly unless this is related to a pattern of dysfunctional, chaotic violence and lifestyles (McKay, 2022).
While there is considerable literature on the nature of domestic violence, and on the structure of parole decisions in general, there is currently no research on parole decisions specifically for perpetrators of domestic violence.

Aims of this research
This research examines how parole boards reach decisions about men who abuse female partners in England and Wales.To do so, the research aims: • To compare variables associated with decisions to release prisoners against variables identified by Parole Board members and explore any differences between actual and perceived decision making.• To explore conscious decision making in domestic violence cases from the perspective of Parole Board members.• To examine the role played by structural factors in parole decision making, in the context of tensions over the Parole Board's role.• To apply decision-making research and the study of heuristics to parole decisions.

METHODS
This research involves a 'mixed-priority' design where qualitative and quantitative elements are equally privileged and feed into each other throughout (Johnson, Onwuegbuzie & Turner, 2007).
A qualitative exploration of the experience of decision makers was important because it could explore the thought process underpinning conscious decision making.This would not, however, provide a complete picture of how decisions are made in practice, given evidence on decision making in general, and the gap between conscious, analytical decision making and unconscious, intuitive decision making.A quantitative analysis of actual parole decisions was therefore needed, to 'triangulate' the qualitative data.

The qualitative element: thematic analysis of interviews with Parole Board members
Twenty semi-structured interviews were carried out, via telephone, with current members of the England and Wales Parole Board.Survivors and practitioners in the field of domestic violence contributed to the interview guide and interviewees were paid for their time.The coding, detailed in an earlier publication (Dyke, Bird & Rivas, 2020), involved a 'reflective' thematic analysis (Braun & Clarke, 2012) with three increasingly-analytical rounds.

The quantitative element: logistic regressions and latent class analyses of a dataset of decision letters
While a previous mixed-methods study of parole decisions in England and Wales used the parole dossier -comprising several hundred pages -as their unit of analysis (Bradford & Cowell, 2012), this study considered only decision letters, a document of five to ten pages written by the Parole Board chair within a fortnight of the hearing (Parole Board, 2019).
The dataset used parole decision letters from April 2018 to September 2019, starting with every decision from that period, filtered to leave those cases where domestic violence was the sole or primary focus.
There were 142 decision letters that met those criteria.Five were excluded that involved psychiatric patients, same-sex violence, a non-cisgender couple, or female-on-male violence.This is not due to lack of concern about violence outside the male-to-female dynamic: domestic violence affects LGBTQ people disproportionately and there are significant gaps in the research on femaleto-male violence (Goldenson et al., 2007).However, the tiny number of cases involving anything other than male-on-female violence prevented any meaningful conclusions about those groups.The final sample was thus 137 cases that involved cisgender males abusing cisgender females in heterosexual relationships, subject to custodial sentences in the general prison population.
The data within the decision letters were coded, distinguishing 'situational couple violence' from 'intimate terrorism' and assigning narrative information to discrete categories.(e.g., the degree to which the board felt that the offender had shown remorse).Since most variables were categorical, and the outcome was categorical (release / progress to open conditions / reject application), logistic regressions, calculating the impact of each variable on the likelihood of release/progression, were best suited to the analysis (Bradford & Cowell, 2012).Latent class analyses grouped cases based on the most significant variables.

Limitations and strengths of this research
This research has significant strengths, as the first known study of parole decisions focusing on domestic violence perpetrators.It uses novel primary data, reaching theme and meaning saturation in the interviews consistent with previous findings on the point at which further interviews yield diminishing returns (Bradford & Cowell, 2012).It makes statistically significant findings from the quantitative data, meeting Lincoln & Guba's (1985) framework for trustworthiness in qualitative research.The study applies behavioural psychology (heuristics and unconscious decision making) to the criminal justice system, while identifying the potential application for signalling theory in the study of DVPPs.It makes findings around the difference between decisions for all crimes and decisions for domestic violence offences.The quantitative findings were buttressed by sensitivity analyses, using alternative delineations to mitigate any subjectivity in how categories had been chosen.These sensitivity analyses produced no differences as to which associations were statistically significant.
The researcher has been an expert witness in the courts with experience of hundreds of domestic violence cases, is familiar with the judicial system and is a male professional examining a topic where professionals assess risk posed by men towards women.A different researcher could have produced different perspectives and different findings from the interviews.The researcher's analyses were, however, discussed with the co-authors who have different positionalities as a survivor-researcher and academic, which helped establish the credibility of the findings.
The study is limited to decisions made in England and Wales in 2018 and 2019, involving cisgender men who abused cisgender women.Caution should be taken in applying the findings beyond that scope.
The results may over-emphasise the perspective of former probation officers.Compared with the full list of 283 current members (Parole Board, 2023), the sample over-represents members with a probation background (40% of interviewees, but only 13% of current members) at the expense of judicial and independent members.However, the interviewees nevertheless included a range of members from judicial, psychological and independent backgrounds.

Overall findings
Logistic regressions identified 17 variables associated with a decision to release, where the probability of the association being a coincidence was below 5% (p < 0.05).These factors are listed in Table 1 along with the variables that did not show significant associations.

The role of professional recommendations
Professional recommendations stand out as the best predictors of release decisions.Of the 137 cases, the Parole Board released just one prisoner over the unanimous objections of the professionals.In that exceptional case, the board overruled an offender manager and offender supervisor who recommended that the prisoner remain in custody.There were no cases in the sample of the Parole Board overruling a negative recommendation from a psychologist, excluding three cases where two psychologists disagreed (so there is no odds ratio in the results).The converse did not apply -while parole boards usually agreed with a recommendation to release, they sometimes did not.The strongest predictors of a release decision were the recommendations of the offender supervisor (odds ratio 255.5, p < 0.001, R 2 = 0.713), psychologist (R 2 = 0.682) and offender manager (odds ratio 94.6, p < 0.001, R 2 = 0.670).
The finding is even more stark when subject to a latent class analysis in Table 2 -prisoners with professional recommendation for release were far more likely to be released than those without, whether using a 2-class model (odds ratio 217.1, p < 0.001, R 2 = 0.673, entropy 0.94) or 3-class model (odds ratio 894.7, p < 0.001, R 2 = 0.774, entropy 0.775).
The result can also be better visualised in Figure 1.
The analysis shows that the professional recommendation acts as a necessary but insufficient condition for release -a negative recommendation acts as an effective veto against release.
Qualitative analysis of the interviews supported this finding -a key theme concerned the role played by professional reports and DVPPs in the decision-making process.Interviewees did not consider these inputs definitive -when asked an open question about the most important   elements that helped them reach a decision, most interviewees did not mention them -but consistently framed the professional recommendation as a 'hook' on which to hang a release decision.
The hook is necessary because of the risk-aversion emerging from the sensitive and pressurised nature of the parole decision (and inherent in the presumption of continued imprisonment in the statutory test).The board is required to consider release where imprisonment is no longer necessary to reduce societal harm.However, there is also pressure to avoid organisational harm, which could occur if a decision to release a prisoner triggered a negative public and political response: Let's assume this decision blows up in my face . . .if I could say in hindsight that I did this, this, and this, then I feel that can make that decision, more confident about the wolves that are out there . . .(interviewee 7, probation officer) An offender completing a DVPP while in prison also represented a 'hook' to allow release, if board members were otherwise confident in the manageability of the prisoner's risk: It gives you a defensible case for making a judgment that says 'yes' in deciding to release someone.(interviewee 3, probation officer) In other words, Parole Board members valued attendance on a DVPP for the same reason that they valued a professional recommendation for release: for its inherent value, and because it offered a concrete basis for them to shift from a risk-averse starting point.

6.3
The role of previous offending Some of the variables that did not produce significant associations are also notable: despite the emphasis in the qualitative interviews on the 'pattern of offending', the offender's previous history of offending and severity of offending were not significant predictors of the parole decision.The only characteristic of the offending that met the statistical significance threshold was the 'nature' of the offending: prisoners were more likely to be released (odds ratio 6.8, p < 0.001, R 2 = 0.164) if their violence lacked a controlling element.Perpetrators of 'situational couple violence' were more likely to be released than those who had demonstrated controlling behaviours common to 'intimate terrorism' (Johnson, 2006).Decisions were also unaffected by whether the offender had reduced their drug and alcohol use since their offence.The qualitative data highlighted the nuanced role of the 'pattern of offending' in the Parole Board's risk assessment of the perpetrator.Every interviewee named the 'pattern of offending' as the first consideration when reaching a decision.However, that pattern of offending proved more subtle than simply an evaluation of the frequency or severity of previous domestic violence.We had already seen that a prisoner with a history of offending or more serious offending was no less likely to be released.Instead, Parole Board members understood the 'pattern' as a more complex concept that interacted with the 'manageability' of the risk.Parole Board members reached decisions to release, based rarely on their confidence that the offender no longer presented a risk, but usually on their confidence that their risk could be managed by probation services and other external factors.
The key to this complexity was the 'nature' of previous offending, and how that nature affects the offender's manageability.Interviewees identified a significant implication of more chaotic offending behaviour: At least if there's a pattern then there's some predictability.(interviewee 6, probation officer) As other interviewees pointed out, it can be far harder to predict the manageability of an offender who has committed one horrific crime (often a murder), whereas a dysfunctional offender whose violent acts followed a familiar pattern of deteriorating behaviour might be easier to managethough their risk of reoffending would be similarly high (maybe higher), a capable probation officer would have more chance of identifying concerning behaviour which suggested an increased risk, and recall them to prison.Controlling behaviour was especially difficult to detect -as multiple interviewees reflected, a controlling perpetrator who stops drinking may become better at maintaining control and avoiding detection, while someone with a habit of getting drunk and as a result inflicting violence on others (including their partner) may reduce their risk through sobriety: I'm usually trying to decide: 'is the risk this person presents a manageable one or not?', which is a way of avoiding the question 'is this person going to do it?',it's subtly different.(interviewee 13, judicial member) This explains why prisoners with a long history of offences may be more likely (not less) to be released: their prior behaviour shows patterns that can be picked up by a capable probation officer.
Consequently, decisions are affected by factors external to the prisoner, such as the professional structure around them: interviewees emphasised the importance of being confident in the risk management plan and in the capability of the probation officer.This professional variable was compounded by an institutional variable: interviewees consistently indicated respect for probation officers but lamented the destabilising impact of the attempted privatisation of the service and the result that many probation officers had no prior knowledge of the case.They could therefore be more likely to release in a marginal case where they were confident in the probation officer and their risk management plan, while the same prisoner with a less robust plan and a new probation officer might see their application rejected, despite posing the same inherent risk.

Other factors influencing the parole decision
As demonstrated in Table 1, the board's positive impression of the offender's honesty and openness (odds ratio 57.2, p < 0.001, R 2 = 0.485), working relationship with professionals (odds ratio 18.7, p < 0.001, R 2 = 0.334), and the effectiveness of the risk management plan (odds ratio 27.1, p < 0.001, R 2 = 0.453), were also significantly associated with a decision to release or progress.Less pronounced but still statistically significant variables included a lower imminence of risk, protective community links, good behaviour in prison and on licence, being subject to an indeterminate rather than determinate sentence and completing a DVPP.Several interviewees also admitted to being more risk-averse when they started hearing cases than after a few years of experience, and they valued the opportunity to discuss marginal cases with colleagues.Some suggested that board members with experience as lawyers or judges might be more inclined to release than those from lay backgrounds.
Parole Board members are conscious of wider contexts: every interviewee mentioned, unprompted, the case of John Worboys, and acknowledged that Parole Board members are aware of public pressure not to release someone who might reoffend, or who is not deemed to have 'paid' for their crimes.

DISCUSSION
The pathway through the Parole Board's decision about a domestic violence perpetrator can be understood to be composed of four key elements: an initial risk-aversion (influenced by external and structural factors); the role of professional witnesses as 'gatekeepers' for release; an evaluation of risk incorporating offence-specific research; and the impact of the interaction between risk manageability and the type of offending.

The 'risk-averse' default position of continued imprisonment
Quantitative findings support previous research (Meyer, 2001;Proctor, 1999) which suggests that parole boards require a case to pass certain hurdles before they can even consider a release, implying a default aversion to release consistent with the statutory test for release, but enhanced by these pressures.The parole decision therefore begins before the hearing, and before the decision makers are aware of the case, due to the statutory test for release (which treats continued imprisonment as the default) but also due to external pressures.These pressures are mostly risk-averse, given popular, media and political pressure against release, and indirect effects of the bureaucratisation of the service (King, 2018).The pressure to reduce not only societal risk (the prospect of a prisoner reoffending after release) but organisational risk (the prospect of the Parole Board suffering damage to its standing due to a contentious release) nudges board members to err on the side of rejecting a parole application, in line with the notion of 'risk colonisation' (Rothstein, Huber & Gaskell, 2006).At a psychological level, this pressure and the degree of uncertainty amplifies the 'status quo bias', especially since decision makers are more likely to regret a decision to change a situation than a decision to maintain the status quo (Kahneman & Klein, 2009).None of this prevents members of a Parole Board from releasing a prisoner -they would not have taken on the role were they not resilient to pressure -but it creates an environment where decision makers need to base a decision to release on something 'concrete', rather than 'only' on their comprehensive evaluation of the risks posed by the prisoner.
The findings support Proctor's (1999) 'new parole' model which highlighted 'eligibility' factors such the nature of the sentence and type of prison.Proctor found these factors more important than case-specific factors, and the dataset showed that prisoners in open prisons or subject to indeterminate (life or IPP) sentences were more likely to be released than those in closed prisons or serving determinate sentences.The case of IPP sentences is the exception to the rule about external pressure promoting risk-aversion: interviewees acknowledged an organisational desire to 'un-stick' prisoners subject to these contentious (Annison & Condry, 2022), open-ended sentences.

7.2
Professional witnesses (especially the psychologist) then act as 'gatekeepers' to the prospect of release The role of professionals, especially the psychologist, is one of 'gatekeepers' in the decision, consistent with previous findings around the 'sequential' model of parole decisions and the need for a professional 'sponsor' (Meyer, 2001) -what Proctor (1999) described as 'institutional factors' -before the board considers release.A recommendation in favour of release 'allows' the Parole Board to move to the next phase of deliberation, but a negative recommendation effectively precludes release.Interviewees felt that psychologists were usually the 'drivers' behind other professionals' recommendations and were uncomfortable with the way psychologists were 'put on a pedestal'.This is consistent with prisoners' own impression that they will not be released unless the psychologist recommends it (Shingler & Needs, 2018;Warr, 2020b).
But while parole boards may not regard the psychologist's view as infallible, they are unavoidably aware of its weight.In marginal cases, they may believe the level of 'societal harm' (Rothstein, Huber & Gaskell, 2006) is lower than the psychologist believes.However, parole decisions carry high stakes, and if the Parole Board overrules a psychologist and prematurely releases an offender who goes on to reoffend, the 'organisational harm' is greater, since a decision to overrule a clinical professional is harder to defend.A change to the status quo (i.e., a release) also requires greater cognitive resources, and holds the greater risk of regret in hindsight (Kahneman, Knetsch & Thaler, 1991), which may also account for the higher likelihood of overruling a recommendation for release than one against release.

7.3
Parole boards take a tailored approach to evaluating domestic violence risk Before considering individual case characteristics, the decision has been influenced by external pressures, professional recommendations, and possibly the composition and backgrounds of the panel, but case-level characteristics nevertheless remain the conscious focus of the decision-making process.
Parole boards are more likely to release domestic violence perpetrators who show: greater 'instrumental compliance' (Bottoms, 2001) evidenced by positive working relationships with professionals; completion of a DVPP; and compliant behaviour in prison and in the community, consistent with previous findings for other groups of offenders (Caplan, 2010;Huebner & Bynum, 2008).
However, there are notable differences.While alcohol and drug use influences parole decisions for crimes in general (Bradford & Cowell, 2012;Mooney & Daffern, 2014), both quantitative and qualitative elements of this study found that Parole Board members placed minimal weight on this factor for perpetrators of domestic violence.This reflects research on the 'blame-it-on-thedrink fallacy' and the importance of not attributing controlling, abusive behaviour to intoxication (Galvani, 2006).
These findings suggest that Parole Board members are attuned to relevant research -while they apply the same test of risk-manageability for all cases, they apply a tailored analysis to domestic violence cases, based on the way risk needs to be evaluated for different types of offenders.This shows that despite the chronic lack of training for Parole Board members, highlighted in the interviews and in the JUSTICE (2022) report, Parole Board members are receptive to offence-specific research evidence and effective at incorporating it into their decisions.
Interviewees consistently framed decisions in intuitive terms, emphasising their professional reaction to the prisoner in person, while lamenting the lack of feedback on how these cases progress -research on intuition versus analysis in decision making has consistently found intuitive judgment more effective when feedback is made available shortly after the decision (Kahneman & Klein, 2009).

7.4
Parole boards treat perpetrator programmes as a 'signal', not a mechanism for change Likewise, the quantitative association between attending DVPPs and an increased likelihood of release does not mean that Parole Board members have confidence in the effectiveness of the programmes.Parole Board members were just as lukewarm about the programmes' effectiveness as existing research suggests they should be (Haggård et al., 2017;Vigurs et al., 2016), but recognised that more insightful and remorseful prisoners were more likely to attend the programme to begin with.A prisoner who refused to attend a DVPP would cause understandable concern about their ability to reflect and change.This suggests an unexpected application of 'signalling theory' (Spence, 1978), usually applied to the economics of programmes of academic study, whereby a programme's value lies in signalling to others that they had the personal resources to complete the course in the first place and that the cost (defined broadly) of attending the course would be lower.
Attendance on programmes is therefore significantly associated with release from prison, but not because parole boards are confident in their power to reduce an offender's risk.Instead, these interviews suggest that they see attendance as a proxy indicator for more important characteristics -the level of insight, remorse and openness shown by the perpetrator.This is supported by the quantitative data showing that DVPP attendance is most correlated with release when combined with the board's positive impression of these characteristics.This echoes concerns from JUSTICE (2022) about overreliance on perpetrator programmes among prisoners and Parole Board members: some prisoners are motivated to change but unable (practically or intellectually) to engage with a programme; some are unmotivated but manipulative; some programmes are ineffective at addressing offending behaviour; and they have become a task to complete rather than a way of achieving change.
The implication here is that DVPPs in their current form are significant to the parole process in a signalling role.This could be as significant for the role of DVPPs as Spence's (1978) findings were for the role of higher education.Policy decisions around DVPPs will be different if the programmes simply filter, rather than transform, violent offenders.The most interesting finding about case-specific characteristics relates to an interaction: between the Parole Board's confidence in the offender manager being able to recall the offender should their behaviour cause concern; and the nature of the offender's crimes.This interaction links the previous work on 'constraint-based compliance' (Bottoms, 2001) and the distinction between 'situational couple violence' and 'intimate terrorism' in domestic violence (Johnson, 2006).
Interviewees consistently said that the 'pattern of offending' was the first thing on their mind when hearing a case, so it was unexpected that the 'nature' of offending ('situational couple violence' rather than 'intimate terrorism') was the only characteristic of the offending associated with a release decision, rather than the frequency or severity of the offending, or whether the prisoner killed or raped his victim.But this apparent contradiction makes sense in the context of risk manageability since a pattern of offending is still a pattern.
Research showing a higher reoffending risk for more severe or more frequent prior offending (Caplan, 2010;Huebner & Bynum, 2008;Mooney & Daffern, 2014) affects the first part of the Parole Board's thought process about offenders: 'is he a lower risk of harm?'.However, interviewees were rarely convinced that an offender was low risk -more often, their decision hinged on a second question: 'can his risk be managed?', and whether the offender manager could monitor them effectively.The perceived quality of the risk management plan was significantly associated with a decision to release a prisoner, echoing the role of 'luck' (Padfield, 2019), whereby two prisoners posing the same risk may get different parole decisions depending on the board's confidence in the probation officer.
While no interviewees explicitly referred to Johnson's (2006) typology, this research suggests a role for the distinction between 'situational couple violence' and 'intimate terrorism' (and for the importance of power and control in domestic violence) in predicting parole decisions.

Psychologists' reports merit particular attention given their effective veto over release
This research highlights the profound (and under-acknowledged) impact of professional recommendations on the parole decision, the elevated role of the psychologist among those professionals, and how these recommendations function as a 'hook' to allow a Parole Board to consider release.This is consistent with previous findings about the importance of an institutional 'sponsor' (Meyer, 2001;Proctor, 1999) while accompanying misgivings about the extensive delays to parole hearings caused by the commissioning of psychologist reports (JUSTICE, 2022).This finding provides context for recent efforts by government (later withdrawn after being deemed unlawful by the courts) to prevent the Parole Board receiving expert recommendations.Since the data collection for this research, the government passed a Statutory Instrument (The Parole Board (Amendment) Rules 2022) creating two significant changes to parole hearings in England and Wales: (i) Hearings could now be held in public.Psychologist reports are especially significant in unpicking the ways in which prisoners must form different identities: a persona that helps them navigate the prison life among fellow inmates; and a 'flagellant persona' that they must present to the Parole Board to plead their case for release, resulting in a 'narrative labour' (Warr, 2020a) as they try to maintain both identities.The role of the Parole Board in considering offenders serving indeterminate sentences amplifies this burden further, since the lifer or IPP prisoner does not know how long they must maintain each identity.A psychologist's input is therefore particularly valuable in such cases.
This finding should also reassure all professionals giving evidence of the significance of their recommendations and the enhanced value of genuinely independent authoritative decisions.Parole boards are far more impressed by professionals who demonstrate knowledge of the case, rigorous planning and logical formulation of their decisions based on their own judgment rather than relying on the psychologist's.
All those involved in the parole hearing, and those researching parole decisions, need to recognise the critical role the psychologists' recommendations play in the decision, and devote proportionate resources to respectively formulating, critiquing and studying how those decisions are reached.Policymakers should also recognise the effective 'veto' psychologists hold over a release, and consider how this affects the structure of the hearing.
This requires honest reflections on the purpose of the psychologist report -whether commissioning these reports is an honest attempt to reduce societal risk (through the insights it provides) or a means (subconscious or otherwise) of reducing organisational risk (Rothstein, Huber & Gaskell, 2006), by providing 'cover' for a release decision.If the latter, then their frequency should be reduced.If the former, then one might question the merits and ethics of holding a hearing at all when the unanimous professional recommendation was opposed to release -formalising that professional 'veto' over release might cause unease, but this may already be happening informally in practice, potentially offering false hope to prisoners.

The size and composition of the panel may affect the decision
While not statistically significant, prisoners were released more often by a panel of three than by a panel of one or two -this information was often absent in the decision letter and needs to be tested with a larger sample.However, interviewees consistently referred to the reassurance of having colleagues next to them, and the isolation of hearing a case alone.
The interviews also suggest that more experienced Parole Board members are more likely to release a prisoner in a marginal case.If true, this infers a hypothesis: that the association between a larger panel and a decision to release might only appear statistically significant when controlled for the experience level of the members -a chair hearing a case on their own may be more riskaverse about release compared with the same chair hearing a case with colleagues, but they are also more experienced than newer members who only sit with one or two colleagues.
If larger panels are more likely to release a prisoner, this suggests that reducing panel size to cut costs -as occurred due to the freeze on recruitment between 2012 and 2016 (JUSTICE, 2022)is self-defeating.The cost of keeping a prisoner incarcerated for several more months (assuming the panel was satisfied, based on the evidence, of their suitability for release) vastly exceeds the cost of paying other Parole Board members to hear the case.This potential finding adds further weight to the JUSTICE (2022) recommendation to remove the presumption that a panel should start with one member.
This research found that intuitive decision making is commonly applied in parole decisions.If the Parole Board provided periodic, short feedback to members on its decisions, it would bring its decision making into line with comparable scenarios where intuitive decision making is more reliable (Kahneman & Klein, 2009): scenarios with 'high validity' where the decision maker knows as soon as possible what happened after their decision.While a Parole Board member can never get the instant, visceral feedback of an A&E surgeon or a firefighter, systematic feedback would improve the current process where they are only likely to hear feedback on cases where the prisoner reoffended.
This research found that Parole Board members were both reliant on intuition in their decision making and keen to absorb theoretical and empirical evidence.Therefore, brief feedback along the following lines should be low cost and expected to improve the quality of decision making: Making this feedback routine would encourage learning from past decisions and normalise awareness of the outcomes of decisions further down the line.The feedback should include feedback on prisoners they did not release, so that Parole Board members became equally conscious of the outcomes for either decision.

8.3
The probation officer's role in a parole decision remains crucial Despite the perception by prisoners that the psychologist's view dominates (Shingler & Needs, 2018) and the findings in this research supporting that view, probation officers should neither feel less important in the parole hearing nor should they be treated as a secondary service.Parole boards are far more confident in releasing a prisoner in a marginal case when they have confidence in the personal and organisational effectiveness of the probation officer assigned to the offender upon release.Under-funding or otherwise undermining the effectiveness of the probation service is therefore a false economy, since the 'constraint-based compliance' (Bottoms, 2001) provided by probation officers keeps prisoners in prison at huge cost when they could otherwise have been managed in the community.There is also an injustice: two identical offenders could have different outcomes from a parole hearing depending on whether they were fortunate enough to have a capable and well-supported offender manager (Padfield, 2019).This finding further highlights the significant societal cost of abortive reorganisations (Carr, 2020) and woefully insufficient resources (JUSTICE, 2022) experienced by the probation service.

Areas for further research
The methods outlined in this study serve as a model for the 'researcher-in-practice', typically a sole researcher balancing research with a full-time professional role, but with enhanced access to data and the opportunity to incorporate findings into practice while embedding practice wisdom in research.Future research should explore how psychologists reach their decisive conclusions, incorporating the powerful impact of the psychologist's views on the prisoner's sense of self and identity (Warr, 2020b).
Research would also be valuable into whether the nature of offending -'situational couple violence' or 'intimate terrorism' (Johnson, 2006) -is associated with different rates of reoffending after release.
Researchers could also consider whether other perpetrator programmes also fulfil a 'signalling' role in the parole decision, and whether the number and background of Parole Board members at a hearing statistically affects the likelihood of release.

CONCLUSIONS
This is the first study to explore parole decision making for domestic violence perpetrators.By studying parole decisions for a single type of offence (domestic violence) and comparing a dataset of actual decisions with the reflections of Parole Board members, several themes become more apparent.Parole boards in England and Wales tailor their analysis of risk to the nature and type of the offence -they are less interested in whether the prisoner has reduced their drug and alcohol use, and more interested in the presence of control and 'intimate terrorism' in their initial offending.The results support previous models which highlight the critical role of professional (namely the psychologist's) recommendations which amount to an effective 'veto' over release.Policy and research around the parole process need to pay special attention to the way psychologists' reports are formulated and incorporated into the hearing, and to the role played by DVPPs, which may be more significant as a 'signal' of an offender's existing risk than as a tool for reducing that risk.

20591101, 0 ,
Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/hojo.12551 by University College London UCL Library Services, Wiley Online Library on [23/01/2024].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 20591101, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/hojo.12551 by University College London UCL Library Services, Wiley Online Library on [23/01/2024].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License TA B L E 1 The relationships between case-specific variables and whether the prisoner was released (or progressed to open conditions), in all 137 parole decisions between April 2018 and September 2019 where domestic violence was the sole or primary consideration.Ordered by effect size (Nagelkerke R 2

20591101, 0 ,
Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/hojo.12551 by University College London UCL Library Services, Wiley Online Library on [23/01/2024].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License F I G U R E 1 Graphical representation of the latent class analyses in Table 2

20591101, 0 ,
Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/hojo.12551 by University College London UCL Library Services, Wiley Online Library on [23/01/2024].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License

20591101, 0 ,
Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/hojo.12551 by University College London UCL Library Services, Wiley Online Library on [23/01/2024].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License

20591101, 0 ,
Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/hojo.12551 by University College London UCL Library Services, Wiley Online Library on [23/01/2024].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License

20591101, 0 ,
Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/hojo.12551 by University College London UCL Library Services, Wiley Online Library on [23/01/2024].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 7.5 'Patterns' of offending can make an offender's risk seem more manageable

20591101, 0 ,
Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/hojo.12551 by University College London UCL Library Services, Wiley Online Library on [23/01/2024].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License (ii) While board members would still receive reports from professionals, the reports should no longer carry a recommendation from the author on whether the offender should be released.While the former received more press attention (BBC News, 2023), this research suggests that the latter had more potential to change the way parole boards reach decisions in England and Wales, due to the critical role played by those expert recommendations.In March 2023, the High Court declared the rule change unlawful (R (Bailey and Morris) v. Secretary of State for Justice (2023)), preventing the removal (at the time of writing) of what turns out to be the most consequential element of the parole decision-making process.
6-month follow-up on decision on [date] to grant release for [Prisoner A].Prisoner recalled on [date] due to breach of licence conditions.or 12-month follow-up on decision on [date] to refuse release for [Prisoner C].Prisoner released on [date] at next hearing; remains in the community on licence.(author's suggested wording) This research is funded by an Economic and Social Research Council stipend for a PhD for Chris Dyke, supervised by the other two authors.The authors are also grateful to the Parole Board Research Governance Group for their permission, and to Glenn Gathercole for his kind assistance in conducting this research.O R C I D Chris Dyke https://orcid.org/0000-0003-4380-4132R E F E R E N C E S Albertson, K. & Corcoran, M. (2020) Marketisation and privatisation in criminal justice.Bristol: Policy Press.20591101,0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/hojo.12551 by University College London UCL Library Services, Wiley Online Library on [23/01/2024].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License